Add libcurl to required packages (partially fixes issue #63)

[stweil]

Checklist

• [x] Used a personal fork of the feedstock to propose changes
• [x] Bumped the build number (if the version is unchanged)
• [ ] Reset the build number to 0 (if the version changed)
• [ ] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• [ ] Ensured the license file is being packaged.

[conda-forge-webservices[bot]]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[stweil]

tesseract with curl can directly process images in the web (by using their URL) without the need of a local copy.

[carlodri]

My two cents:

• I think that this will not work unless the appropriate configure options are used
• could we test libcurl in the tests with some command using a URL?

[stweil]

• configure checks for available libraries (using pkg-config) and uses available ones by default, so typically no configure options are required
• configure --with-archive --with-curl will fail if the required libraries are not available
• run tesseract 'https://github.com/tesseract-ocr/test/blob/main/testing/HelloGoogle.tif?raw=true' - and expect the result Hello Google

[scw]

I think building with --with-archive --with-curl is best for consistency with the Windows builds and makes explicit the dependencies that will be picked up from the host section.

[stweil]

I think building with --with-archive --with-curl is best for consistency with the Windows builds [...]

This would work for Windows and Linux, but currently fail for macOS where libarchive is not found.

[scw]

This would work for Windows and Linux, but currently fail for macOS where libarchive is not found.

OK, I did some digging on this. This is because libarchive on MacOS depends on libiconv, which the package correctly resolves as a dependency, but currently libiconv does not expose pkg-config information. This is logged as https://github.com/conda-forge/libiconv-feedstock/issues/36. Manually running the command ./configure calls of pkg-config --exists --print-errors "libarchive" shows:

./pkg-config --exists --print-errors "libarchive"
Package iconv was not found in the pkg-config search path.
Perhaps you should add the directory containing `iconv.pc'
to the PKG_CONFIG_PATH environment variable
Package 'iconv', required by 'libarchive', not found

I think the better fix is to address libiconv, but I did confirm that locally changing the command to:

libarchive_CFLAGS="-I${CONDA_PREFIX}/include" libarchive_LIBS="-I${CONDA_PREFIX}/lib" ./configure --with-archive --with-curl

Was able to address the issue as a workaround if you'd like to take that tack until libiconv is addressed:

checking for pkg-config... /usr/local/miniconda/envs/tesseract-build/bin/pkg-config
checking pkg-config is at least version 0.9.0... yes
checking for libcurl... yes
checking for lept >= 1.74... yes
checking for libarchive... yes
checking for icu-uc >= 52.1... yes
checking for icu-i18n >= 52.1... yes

[stweil]

Thanks. I added your suggestions in a 2nd commit. Now the macOS builds use libarchive, too.

[scw]

@stweil could you try again with $PREFIX instead of $CONDA_PREFIX? I had run that locally on an environment rather than via conda-build.

[stweil]

@stweil could you try again with $PREFIX instead of $CONDA_PREFIX? I had run that locally on an environment rather than via conda-build.

Now all builds fail with a linker error because of a missing symbol from libarchive.